Peter Krautzberger · on the web

MathML Forges On -- notes, AAP leftovers, and a summary

"The end (tm)" of MathML in Chrome this week happened to coincide with a piece I had been working on for a while now and which was published at O'Reilly's Programming blog ("MathML Forges On") yesterday. Originally, I had meant to write something for O'Reilly's TOC blog earlier this year, but then the TOC conference was retired (and so was the blog) so things got delayed.

Then in the summer, the ever fantastic Tzviya Siegmann pulled me into the AAP's EPUB implementation project and so I had a chance to be active in both the features and the accessibility groups. It was an extremely interesting experience all around so just two points. First, I am amazed at the work of the group leaders, handling the difficulties of running such a complex project, in a very short time, with a highly diverse group of participants. Second, it was an experience to, well, experience anti-trust concerns like that. While frustrating at times, I didn't find this tragic in the end; there was more to lose than to gain. But it felt strange to be on the other end of the stick, so to speak, wanting to drive publishers and reading systems to close collaboration so that we can finally get decent support for math & science in ebooks. The AAP white paper came out this week.

During that time I also wrote the first drafts for "MathML Forges On" (the title was a suggestion of Simon St. Laurent who has been a great editor). I want to thank Fred Wang, David Carlisle, Neil Soiffer and Dave Barton for many helpful comments and Sanders Kleinfeld for being a matchmaker, twice.

During the AAP project, I also summarized the technical details a little with suggestions for the short term. This wasn't useful to the AAP project but there's no reason to throw it away -- so here it is.

A sort of summary

Note. This was written with the following question in mind: how can we speed up MathML adoption in epub3? Well, one way would be to understand what level of support can be achieved in the short term. Since full MathML 3 support is simply not available, compromises have to be made. This means creating guidelines for publishers to ensure their content can be supported and for reading systems to understand how they can support that level. And for both sides to push each other -- publishers, pushing reading systems by pointing out an achievable level of support; reading systems pushing publishers to create reasonable MathML.

Native MathML support in browser is limited to partial support in Firefox and Safari; JavaScript polyfills are available and have been used in reading systems.

Recommendation: "Firefox 24"-level MathML support is a good baseline in the short term (6-12 months). This amount of MathML support covers most publishing needs and any ePub3 reading system can realistically provide this level of support via Gecko or polyfills within 6 months. The open source STIX and Asana fonts should be the default math fonts since technical limitations make them the only widely supported fonts in web environments.

Browser support

Polyfill support

MathJax is modular, its primary output options are HTML and SVG, where it implements the TeX layout algorithm. MathJax can also augment native MathML output in Safari and Firefox (trading speed for layout quality). MathJax offers accessibility features (zoom, scale, copy&paste) and works well with existing math accessibility solutions (MathPlayer, ChromeVox). MathJax has been integrated into epub3 reading systems such as Readium, Vitalsource Bookshelf, Azardi and IDEAL reader.

Jqmath tweaks browser layout instead of implementing a separate algorithm, trading speed for layout quality. Jqmath tries to be font agnostic and assumes browsers can access the necessary fonts and unicode points. Since its developer contributed to WebKit's MathML support, it works particularly well on Safari (augmenting the MathML output).


Fonts are a particularly complex issue for mathematics and MathML. Here are a few problems: